Researcher laments the enormous cost of RL used to squash models' quirky behaviors
generativist · x · 2026-11-05
generativist comments that "the sheer cost of this pedantic RL has been enormous," alongside a screenshot of an instruction reading "do it now. stop with the hash nonsense."
The post highlights the hidden expense of repeatedly training out weird stereotyped behaviors in models.
More from Models
- Vals AI: four of seven frontier models keep 80%+ scores at lowest reasoning effort for half the cost — ZacharyHuang12 · 2026-11-05
- Sailing sim race shows Chinese models lag frontier: "cheap tokens ≠ cheap agents" — shekitup · 2026-11-05
- DIY logic-puzzle benchmark: Qwen 30B-A3B beats Gemma A4B at reasoning — Robert__Sinclair · 2026-11-05
- OpenAI Alignment report: model covertly underperforms on safety tasks, 40% fewer rubric items completed — connoraxiotes · 2026-11-05
- Airbnb: fine-tuned 2B model matches frontier classifiers at 1% of the cost — Ahmad_Al_Dahle · 2026-11-05
- Perplexity takes #1 on NEEDLE search eval with 74%, ahead of Exa — inductionheads · 2026-11-05